multi-period performance persistence analysis of hedge funds
TRANSCRIPT
Multi-Period Performance Persistence Analysis of Hedge Funds
Vikas AgarwalNarayan Y. Naik*
Current Version: March 2000
JEL Classification: G10, G19
____________________________________________* Vikas Agarwal, London Business School, Sussex Place, Regent's Park, London NW1 4SA, UnitedKingdom, Tel: +44-171-262 5050, extension 3535; Fax: +44-171-724 7875; email: [email protected] Narayan Y. Naik, London Business School, Sussex Place, Regent's Park, London NW1 4SA, UnitedKingdom, Tel: +44-171-262 5050, extension 3579; Fax: +44-171-724 3317; email: [email protected] would like to thank Ravi Bansal, Stephen Brown (the editor), Jennifer Carpenter, Elroy Dimson,William Goetzmann, David Hsieh, Frans de Roon, Henri Servaes, an anonymous referee, FauchierPartners Ltd. and participants at the hedge fund conference at the Duke University and Issues inPerformance Measurement workshop at European Institute for Advanced Studies in Management,Brussels for many helpful comments and constructive suggestions. Naik is grateful for funding fromInquire UK and the European Commission's TMR program (network ref. ERBFMRXCT 960054). VikasAgarwal is grateful for the financial support from British Council’s Chevening scholarship, EdwardJones’, Frank Russell’s and Fauchier Partner’s scholarships during past three years in the PhDprogramme. We are grateful to Bob Potsic of Hedge Fund Research Inc., Chicago for providing us withthe data. We are responsible for all errors.
Multi-Period Performance Persistence Analysis of Hedge Funds
Abstract
Since hedge funds specify significant lockup periods, we investigate persistence in the
performance of hedge funds using a multi-period framework in which the likelihood of
observing persistence by chance is lower than that in the traditional two-period
framework. Under the null hypothesis of no manager skill (no persistence), the theoretical
distribution of observing wins or losses follows a binomial distribution. We test this
hypothesis using the traditional two-period framework and compare the findings with the
results obtained using our multi-period framework. We examine whether persistence is
sensitive to the length of return measurement intervals by using quarterly, half-yearly and
yearly returns. We find maximum persistence at quarterly horizon indicating that
persistence among hedge fund managers is short-term in nature. It decreases as one moves
to yearly returns and this finding is not sensitive to whether returns are calculated on a
pre- or post-fee basis suggesting that the intra-year persistence finding is not driven by the
way performance fees are imputed. The level of persistence in the multi-period framework
is considerably smaller than that in the two-period framework with virtually no evidence of
persistence using yearly returns under the multi-period framework. Finally persistence,
whenever present, seems to be unrelated to whether the fund took directional bets or not.
1
Multi-Period Performance Persistence Analysis of Hedge Funds
I. Introduction
It is now well known that the traditional active strategies such as investing in mutual
funds, on average, underperform passive investment strategies. The few mutual fund
managers who successfully beat the passive strategies tend to move into the arena of
“alternative” investments and start their own hedge funds. Hedge funds seek to deliver
high absolute returns and typically have features such as hurdle rates, incentive fees with
high watermark provision which help in a better alignment of the interests of managers and
the investors. This has caused many investors following traditional active-passive
strategies to seriously consider replacing the traditional active part of their portfolio with
alternative investment strategies.
The inclusion of hedge funds in a portfolio can potentially result in better risk-return
tradeoffs due to the low correlation between hedge fund returns and the returns on the
traditional asset classes like equities, bonds, currencies, etc. (see Fung and Hsieh (1997),
Agarwal and Naik (2000a)). However, question arises as to whether hedge funds are able
to add value consistently? This is an important issue in the context of hedge funds
because unlike the traditional mutual funds, investment in hedge funds involves a
significant lockup period. This implies that the investors need to have sufficient
information about the performance of hedge funds over a long period before committing
2
their money to hedge funds1. Moreover, as hedge funds exhibit a much higher attrition rate
compared to mutual funds (see Brown et. al. (1999) and Liang (2000)), the issue of
performance persistence becomes especially important in the case of hedge funds.
This paper contributes to our understanding of persistence among hedge funds in two
important ways. First, it examines whether the nature of persistence in the performance of
hedge funds is of short-term or long-term in nature. Our understanding of the nature of
persistence among hedge funds is largely due to Brown et al (1999) who use annual
returns of offshore hedge funds. They find virtually no persistence in their sample. In
contrast, this paper employs a new database covering offshore as well as onshore hedge
funds, and examines persistence using high frequency data over longer time period. It is
possible that hedge fund managers exhibit differential degree of persistence at different
return horizons, an issue investigated to some extent in mutual funds literature2.
Therefore, we examine both short-term and long-term persistence in the performance of
hedge funds by investigating their pre-fee and post-fee returns over quarterly, half-yearly
and yearly intervals3.
1 As hedge funds are refrained from advertising about their own performance, the investors have to
generate their own information and conclusions about whether the funds that have done well in the past
will continue to do so in the future.
2 For performance persistence studies in the mutual fund literature, see Brown and Goetzmann (1995),
Carhart (1997), Elton, Gruber and Blake (1996), Goetzmann and Ibbotson (1994), Grinblatt and Titman
(1989, 1992), Gruber (1996), Hendricks et al (1993) and Malkiel (1995).
3 Given the limited history of hedge fund returns, it is not possible to examine 3-5 yearly performance as
done in the case of mutual funds.
3
Second, unlike the existing literature, which restricts attention to performance over
two consecutive periods, we also study persistence by examining the series of wins and
losses for two, three and more consecutive time periods. This allows a direct examination
of extent of multi-period persistence, which is essential before locking up investment over
significantly long periods of time. Under the null hypothesis of no manager skill (which
implies no persistence), the probability of winning and losing in each period equals a half
and is independent of the return horizon. We test this null hypothesis for different hedge
funds individually and collectively over two, three and more consecutive periods. Since the
likelihood of observing a series of wins or losses due to chance is much less than
observing two consecutive wins or losses in a two-period framework, the multi-period
framework is able to discriminate better between persistence due to chance and
persistence due to manager skill. We compare and contrast the findings from our multi-
period analysis with those obtained from the traditional two-period analysis on a pre-fee
and post-fee basis.
We conduct this investigation using data provided by Hedge Fund Research Inc.
(henceforth HFR) which covers returns earned by hedge funds from January 1982 to
December 1998 period. HFR data set provides information about hedge funds both living
and dead, and is known to have lower attrition rate compared to other databases such as
TASS (see Liang (2000)). The lower attrition rate in HFR suggests that it includes fewer
number of funds that fail as compared to other databases. This potentially exacerbates
survivorship bias related problem in studies that employ HFR database. We try to mitigate
4
the problem of spurious inferences caused by survivorship related issues a la Brown et. al.
(1992, 1999) by including data on both, the living as well as the dead hedge funds. They
show that survival induced persistence “anomalies” are mitigated, at least in part, by the
use of appraisal ratio. We therefore examine persistence using alphas as well as appraisal
ratios.
Using net-of-fee returns, we find that the extent of persistence is highest at the
quarterly horizon and decreases as we move to yearly horizon. This continues to remain
true with pre-fee returns as well suggesting that our finding of intra-year persistence is not
driven by the imputation of performance fee. It is important to note that even if there
exists some persistence at the quarterly level, it would be difficult for investors to take
advantage of it due to significantly long lockup periods. It is also important to bear in
mind that fact that most hedge funds only put out audited returns on an annual basis – so
some of the apparent intra-year persistence may be caused by stale valuations. In any
case, persistence at the quarterly but not at the annual horizon among the hedge funds
stands in sharp contrast to the results in Hendricks et al (1993) who find that in mutual
funds, persistence is highest at the two year horizon. We observe that the level of
persistence in the multi-period framework is considerably smaller than that in the two-
period framework. Finally, we find that persistence, whenever present, is unrelated to the
type of strategy (directional or non-directional) followed by the fund.
Rest of the paper is organized as follows. Section II provides the sample description
and classifies it into directional and non-directional hedge fund strategies. Section III
5
describes how pre-fee returns are computed and examines the persistence on a pre-fee and
post-fee basis in the traditional two-period framework using both parametric and non-
parametric techniques. Section IV tests for multi-period persistence by comparing the
observed frequency distribution of wins and losses against a theoretical distribution under
the null hypothesis of no persistence. Section V concludes with suggestions for future
research.
II. Classification of Hedge Funds
Although the term ‘hedge fund’ originated from the equally long and short strategy
employed by managers like Alfred Winslow Jones, the new definition of hedge funds
covers a multitude of different strategies. Unlike the traditional investment arena, since
there does not exist a universally accepted norm to classify the different strategies, we
segregate them into two broad categories: ‘Non-Directional’ and ‘Directional’. Hedge
fund strategies exhibiting low correlation with the market are classified as non-directional
(also commonly referred to as market-neutral), while those having high correlation with
the market are classified as directional4. We further divide these two main categories into
ten popular strategies (see Appendix A for details) and examine persistence in the
performance of hedge funds following each of these strategies5.
4 Note that the non-directional strategies are neutral only to the first moment, i.e., expected returns. They
are not neutral to the second moment, as in volatile periods convergence is not always obtained and
arbitrage based strategies can make losses.
5 Agarwal and Naik (2000b) find that these strategies exhibit significantly different risk exposures towardsdifferent asset class factors, and these risk exposures are broadly consistent with their stated investment
6
Table I provides the description of the sample in terms of the number of funds in each
of the strategies, the time period spanned by each strategy, the number of dead funds
during the sample period, average and median number of funds per period for each hedge
fund strategy. Since the incentive fee is typically worked out based on calendar year
return, we select January to December period for computing annual returns, January-June
and July-December for computing semi-annual returns, and January-March, April-June,
etc. for computing quarterly returns. For our investigation using quarterly, half-yearly and
yearly data, we use returns of 746, 716 and 586 hedge funds respectively spanning the
period of January 1982 to December 1998. In general, as we increase the investment
horizon the number of funds within a particular strategy decreases. This is primarily
because the funds need to have returns for at least two periods before they can be included
in the sample.
We select the first complete period (quarter, half-year or year) after the birth of a fund
for the purpose of our investigation. In case of death of a fund, we include returns till the
end of the fund. We have 27 (15 and 13) dead funds out of a total of 746 (716 and 586)
funds using quarterly (half-yearly and yearly) returns. The attrition rate, defined as the
percentage of dead funds in the total number of funds, is 3.62%, 2.10% and 2.22% using
quarterly, half-yearly and yearly returns, which is consistent with an average annual
attrition rate of 2.17% in HFR database reported by Liang (1999) during 1993-97. This
attrition rate is much lower than the annual attrition rate of about 14% for offshore hedge
objectives.
7
funds during 1987-96 reported by Brown, Goetzmann and Ibbotson (1999) and 8.3% in
TASS database during 1994-98 as reported by Liang (1999).
Having described the sample period and characteristics, we now proceed with the
examination of persistence in the traditional two-period framework using parametric and
non-parametric tests.
III. Parametric and Non-parametric Tests of Persistence
It is well known that different hedge fund strategies involve significantly different risk-
return tradeoffs. Therefore, it may not be prudent to compare the performance of a hedge
fund manager following a given strategy with another manager following a different
strategy. We know from Brown et al. (1999) that the existence of ‘style factor’ can lead to
reversals in the persistence phenomenon because of the differences in the levels of
systematic risk across managers. This is especially relevant in the case of hedge funds,
which are exposed to significantly different levels of risk depending on whether they
follow directional or non-directional strategies6. We, therefore, examine the issue of
performance persistence within individual hedge fund strategies. Specifically, we compare
6 See Brown and Goetzmann (1995) for the importance of relative risk adjustment. They find that the
relative risk-adjusted performance of mutual funds persists from year to year but the absolute performance
measured by alphas does not. In a recent working paper, Agarwal and Naik (2000c) estimate the alpha
against a comprehensive benchmark consisting of passive and option-based strategies.
8
the return of a hedge fund following a particular strategy with the average return earned
by all the hedge funds pursuing that strategy.
We follow Brown et al (1995, 1999) and compare the performance measures in the
current period on the performance measures in the previous period. We employ two
performance measures: the alpha and the appraisal ratio. We define alpha as the return of a
hedge fund using a particular strategy minus the average return for all hedge funds
following the same strategy. It is well known that different hedge funds employ different
degrees of leverage to scale up their alphas7. However, this also scales up the volatility of
their returns - a fact that may not be captured by just looking at the alphas. Therefore, we
also use a second measure called the appraisal ratio. This is defined as the alpha divided by
the residual standard deviation resulting from a regression of the hedge fund return on the
average return of all the hedge funds following that strategy. The appraisal ratio accounts
for the differences in the volatility of returns and is leverage-invariant. Therefore, we also
use the appraisal ratios to investigate the extent of persistence in the performance.
To investigate the issue of persistence in two-period framework, we use regression-
based (parametric) and contingency-table-based (non-parametric) methods. We conduct
all the tests at quarterly, half-yearly and yearly intervals using alphas as well as appraisal
ratios. For the regression-based parametric method, we regress the alphas (appraisal
ratios) during current period on the alphas (appraisal ratios) during the previous period. A
7 This effect has been shown analytically by Park and Staum (1998).
9
positive significant slope coefficient on past alpha (appraisal ratio) suggests that a hedge
fund that did well in a given period did well in the subsequent period and vice-versa.
For the non-parametric method, we construct a contingency table of winners and
losers where a fund is a winner if the alpha of that fund is greater than the median alpha of
all the funds following the same strategy in that period otherwise it is a loser. Persistence
in this context relates to the funds that are winners in two consecutive periods (quarterly,
half-yearly or yearly as the case may be) denoted by WW, or losers in two consecutive
periods, denoted by LL. Similarly, winners in first period and losers in the second period
are denoted by WL and LW denotes the reverse. In this framework, we use both cross-
product ratio (CPR) and Chi-square statistic to detect persistence. CPR defined as
(WW*LL)/(WL*LW), captures the ratio of the funds which show persistence in
performance to the ones which do not. The null hypothesis in this setting represents lack
of persistence for which the CPR equals one. In other words, when there is no persistence,
one would expect each of the four categories denoted by WW, WL, LW and LL to have
25% of the total number of funds. We determine the statistical significance of the CPR by
using the standard error of the natural logarithm of the CPR given by (see Christensen
(1990))
2)ln(
1111
LLLWWLWWCPR +++=σ
We also conduct a Chi-square test comparing the observed frequency distribution of WW,
WL, LW and LL for each hedge fund with the expected frequency distribution. In a recent
paper, Carpenter and Lynch (1999) study the specification and power of various
10
persistence tests. They find that the Chi-square test based on the number of winners and
losers is well specified, powerful and more robust to the presence of survivorship bias
compared to other test methodologies. In our study, we aggregate combinations of
winners and losers (WW, WL, LW and LL) across ten different hedge fund strategies. We
compute the Chi-square statistic as
(WW-D1)2/D1+(WL-D2)2/D2+(LW-D3)2/D3+(LL-D4)2/D4 where
D1 = (WW+WL)*(WW+LW)/N, D2 = (WW+WL)*(WL+LL)/N,
D3 = (LW+LL)*(WW+LW)/N and D4 = (LW+LL)*(WL+LL)/N.
We test this statistic at 5% level corresponding to the critical value of Chi-square statistic
of 3.84 corresponding to the Chi-square distribution with one degree of freedom.
Since fees are imputed, but not paid intra-year, such imputation can potentially
influence persistence measure at the quarterly and half-yearly horizons8. We therefore
conduct persistence tests on a pre-fee basis as well. Towards that end, we estimate the
performance fee paid to each fund at the end of each year based on the fee schedule,
hurdle rate and high watermark provision9. We add back a twelfth of this each month for
the past year to arrive at the pre-fee returns. We repeat all our tests with pre-fee returns
and contrast the findings with those observed using post-fee returns reported in the HFR
database.
8 We thank the referee and the editor for bringing this to our attention.
9 Out of a maximum of 746 funds we use for this study, 616 have a high watermark provision and 119
have a hurdle rate. Hurdle rate is typically the T-Bill or the Eurodollar rate. We also adjust for the
management fee that ranges from 1% to 2%.
11
We conduct the parametric and non-parametric tests for each hedge fund strategy
separately. For the overall persistence results, we aggregate the information on all hedge
funds in each time period. For the sake of brevity we report in Table II the percentage of
cases where statistically significant persistence was observed in each hedge fund strategy
on pre-fee basis and on post-fee basis10. These results are based on both alphas (see Panel
A) and appraisal ratios (see Panel B) computed using quarterly, half-yearly and yearly
returns. We find that, in general, the regression based parametric tests indicate a greater
extent of persistence compared to the non-parametric (CPR and Chi-square) tests.
Interestingly, the Chi-square test which Carpenter and Lynch (1999) find to be well
specified, powerful and more robust indicates higher extent of persistence as compared to
that observed with tests based on CPR. We also find that the extent of persistence is
sensitive to the return measurement interval. In particular, persistence decreases as the
return measurement interval increases. Finally, the extent of persistence does not seem to
be related to whether the fund took directional bets or followed an arbitrage based
strategy.
Brown et al (1999) examine persistence among offshore hedge funds using annual
returns. They consider the possibility that performance persists on a pre-fee basis and that
managers can extract their full value-added through fees. To test this proposition, they
compare persistence on a pre-fee basis with that on a post-fee basis and find similar
results. Our results on an annual basis using both onshore and offshore funds over a
10 A Z-statistic of 1.96 corresponds to significance at 5% level.
12
longer time period confirm this finding. Interestingly, we find that extent of persistence at
quarterly and half-yearly level is higher on a pre-fee basis compared to that observed on a
post-fee basis which is consistent with the possibility suggested by Brown et al (1999)
mentioned above. However, since we continue to observe a comparable level of
persistence on a pre-fee and post-fee basis at quarterly and half-yearly intervals, it suggests
that the intra-year persistence observed on a post-fee basis is not driven by the way the
performance fee is imputed.
Having examined the extent of persistence in a two-period framework, we proceed
with multi-period analysis.
IV. Multi-Period Tests of Persistence
In this section, we extend our investigation from the traditional two-period
framework to a multi-period framework. Towards that end, we construct a series of wins
and losses for each hedge fund and compare the observed frequency distribution with the
theoretical frequency distribution of two and more consecutive wins and losses. For
example, under the null hypothesis of no persistence, the theoretical probability of
observing WWW and LLL equals one-eighth while that of observing WWWW and LLLL
equals one-sixteenth, and so on. We illustrate this using annual returns in Chart I that
shows the theoretical and observed frequency distributions of consecutive wins and losses
13
of non-directional strategies, directional strategies and the overall sample based on alphas
and appraisal ratios11.
We employ the two-sample Kolmogrov-Smirnov (K-S) test to check if the
observed distribution of wins and losses is statistically different from the theoretical
distribution. We report the results of the K-S test in Table III based on alphas and
appraisal ratios in Panels A, B and C for quarterly, half-yearly and yearly post-fee returns
respectively. We show in bold face (italic face) the cases where we find persistence
significant at 5% (10%) level.
Table III highlights three interesting features. First, the extent of persistence
decreases as the return measurement interval increases. For example, at 5% level of
significance, there are four cases of persistence in losers and one case of persistence
among winners based on quarterly appraisal ratios. When we increase the return interval
to half-year, we find only one case of persistence in losers and none among winners while
with yearly return interval, there is no evidence of persistence in either winners or losers.
Second, whenever some persistence is observed, it seems to be driven more by losers than
by winners. This is similar to our earlier findings in the case of the two-period framework.
Once again, directional and non-directional funds seem to exhibit similar degree of
11 We repeat this with quarterly and half-yearly returns as well. We find that the best performance
corresponds to 21, 12 and 9 consecutive wins based on quarterly, half-yearly and yearly returns while the
corresponding numbers for worst performance are 22, 18 and 12 consecutive losses. This suggests that
there are a few very good managers and a few very poor managers.
14
persistence. Finally, the level of persistence based on multi-period performance measure is
considerably smaller than that observed under a two-period framework with no evidence
of persistence at the yearly return horizon even at the 10% level. This is because, unlike
the traditional two-period test, our multi-period test involves tracking the history of series
of successes and failures of individual hedge funds throughout the sample period. This
significantly reduces the likelihood of observing large number of consecutive wins or
losses due to chance factor and therefore has more power to discriminate between the
chance and the skill factors.
Since for large samples the binomial distribution can be approximated by normal
distribution, we conduct a Kolmogorov-Smirnov test comparing the distribution of
consecutive wins and losses of hedge funds with a normal distribution. As before, we
conduct this test based on alphas and appraisal ratios separately for quarterly, half-yearly
and yearly post-fee returns. We report the results in Table IV. Persistence in this
framework is captured by the observed distribution being significantly different from a
normal distribution.
Overall the results exhibit somewhat higher level of persistence than that observed
in Table III. However, we continue to observe the same interesting features. First, the
extent of persistence decreases as the return measurement interval increases. Second,
whenever persistence is observed, it is mainly attributable to losers continuing to be losers.
However, we find evidence of a few good managers who consistently outperform their
peers over long periods indicating the importance of manager selection exercise in the
15
context of hedge funds. Third, both non-directional and directional funds exhibit similar
degree of persistence. Finally, the level of persistence based on multi-period performance
measure is considerably smaller than that observed under a two-period framework with
virtually no evidence of persistence at the yearly return horizon.
We repeat these multi-period tests using pre-fee returns and find virtually identical
results12. For the test reported in Table III, in case of quarterly pre-fee returns, we find 7
cases of significance (at 5% level) as compared to 5 cases with post-fee returns. For half-
yearly and yearly pre-fee (post-fee) returns, the number of significant cases is 4 (3) and 0
(0) respectively. The corresponding number of significant cases for the test reported in
Table IV are 19 (20), 6 (7) and 0 (0) for quarterly, half-yearly and yearly pre-fee (post-
fee) returns respectively. In general, similar to the two-period tests, the extent of
persistence is marginally higher with pre-fee returns compared to post-fee returns.
V. Concluding Remarks
This paper investigated the extent of pre- and post-fee performance persistence
exhibited by hedge funds during January 1982 to December 1998 using the traditional
two-period framework and contrasted the findings with those observed using a multi-
period framework. It also examined whether the persistence observed was sensitive to
whether returns were measured over quarters (short-horizon) or over years (long-
horizon). This is particularly important in the case of hedge funds that specify significant
12 These results are available from authors upon request.
16
lockup periods. Finally, it also investigated whether the way in which the performance fee
is imputed affects the degree of persistence observed among hedge funds.
It found three interesting patterns using both pre-fee and post-fee returns. First, there
exists considerable amount of persistence at quarterly horizon. It reduces as one moves to
yearly returns indicating that persistence among hedge fund managers is primarily short-
term in nature. This is in sharp contrast to the findings in the mutual fund literature, which
show that two years is about the horizon of persistence. However, it is important to bear
in mind the fact that hedge funds stipulate significant lockup periods. This may make it
difficult for investors to take advantage of the short-term persistence observed in the
data13. Second, persistence does not seem to be related to the type of strategy followed by
the fund, i.e., both directional and non-directional funds exhibited similar degree of
persistence. Finally, the level of persistence observed in a multi-period framework is
considerably smaller than that observed under the traditional two-period framework, with
virtually no persistence at the yearly return level in the multi-period framework.
Performance persistence over long periods is an important area of future research.
Since entry and exit into active management involves non-trivial costs and since learning
about manager skill takes time, selecting the right manager becomes a very important
issue. This is especially so in case of hedge funds as they specify significant lock-up
periods. As avenues of future research, it would be interesting to examine whether
13 Our finding of short-term persistence may be attributable, to some extent, to stale valuations resulting
from annual reporting of audited statements by hedge funds.
17
performance persistence is related to characteristic features of hedge funds such as size,
lockup period, incentive fees, etc.14. It would also be interesting to see whether multi-
period analysis of mutual funds exhibits significantly different patterns compared to the
ones observed with hedge funds.
14 Recently, Ackermann, McEnally & Ravenscraft (1999) study the relationship between characteristic
features and performance of hedge funds and find that incentive fees explains the higher performance
partly.
18
References
Ackermann, Carl, Richard McEnally, and David Ravenscraft. “The Performance of Hedge
Funds: Risk, Return and Incentives.” Journal of Finance, 54 (June 1999), 833-874.
Agarwal, Vikas, and Narayan Y. Naik. “On Taking the Alternative Route: Risks,
Rewards, and Performance Persistence of Hedge Funds.” Journal of Alternative
Investments (Spring 2000a).
Agarwal, Vikas, and Narayan Y. Naik. “Generalised Style Analysis of Hedge Funds.”
Journal of Asset Management (April 2000b).
Agarwal, Vikas, and Narayan Y. Naik. “On Benchmarking of Hedge Funds with Passive
and Option-based Strategies.” IFA Working Paper no. 300, (March 2000c), London
Business School.
Brown, Stephen J., William N. Goetzmann. “Performance persistence.” Journal of
Finance, 50 (June 1995), 679-698.
Brown, Stephen J., William N. Goetzmann, and Roger G. Ibbotson. “Offshore hedge
funds: Survival & performance 1989-95.” Journal of Business, 72 (January 1999), 91-
117.
Brown, Stephen J., William N. Goetzmann, Roger G. Ibbotson, and Stephen A. Ross.
“Survivorship bias in performance studies.” Review of Financial Studies, 5 (1992),
553-580.
Carhart, M. “On Persistence in Mutual Fund Performance.” Journal of Finance, 52
(March 1997), 57-82.
19
Carpenter, Jennifer N., and Anthony W. Lynch. “Survivorship Bias and Attrition Effects in
Measures of Performance Persistence.” Journal of Financial Economics, 54
(December 1999), 337-374.
Christensen, Ronald. “Log-Linear Models.” Springer-Verlag, New York (1990).
Elton, Edwin, Martin Gruber, and Christopher Blake. “The persistence of risk-adjusted
mutual fund performance.” Journal of Business, 69 (April 1996), 133-157.
Fung, William, and David A. Hsieh. “Empirical characteristics of dynamic trading
strategies: The case of hedge funds.” The Review of Financial Studies, 10 (Summer
1997), 275-302.
Goetzmann, W.N., and Ibbotson R.G. “Do winners repeat?.” Journal of Portfolio
Management, 20 (Winter 1994), 9-18.
Grinblatt, M., and Titman S. “Mutual fund performance: An analysis of quarterly portfolio
holdings.” Journal of Business, 62 (July 1989), 393-416.
Grinblatt, M., and Titman S. “The persistence of mutual fund performance.” Journal of
Finance 42 (December 1992), 1977-1984.
Gruber, M.J. “Another puzzle: The growth in actively managed mutual funds.” Journal of
Finance, 51 (July 1996), 783-810.
Hendricks, Darryll, Jayendu Patel, and Richard Zeckhauser. “Hot hands in Mutual Funds:
Short-Run Persistence of Relative Performance 1974-88.” Journal of Finance, 48
(March 1993), 93-130.
Liang, Bing. “Hedge funds: The Living and the Dead”, this issue Journal of Financial and
Quantitative Analysis, September, 2000.
20
Malkiel, B. “Returns from investing in Equity Mutual Funds 1971 to 1991.” Journal of
Finance, 50 (June 1995), 549-572.
Park, James M., and Jeremy C. Staum. “Performance Persistence in the Alternative
Investment Industry.” Working Paper (1998), Paradigm Capital Management, Inc.
21
Appendix A
Non-directional Strategies: These strategies do not depend on the direction of any
specific market movement and are commonly referred to as ‘market neutral’ strategies.
These are usually designed to exploit short term market inefficiencies and pricing
discrepancies between related securities while hedging out as much of the market
exposure as possible. Due to the reduced liquidity inherent in many such situations, they
frequently run smaller pools of capital than their counterparts following directional
strategies. Included in this group are the following strategies:
1. Fixed Income Arbitrage - A strategy having long and short bond positions via cash or
derivatives markets in government, corporate and/or asset-backed securities. The risk
of these strategies varies depending on duration, credit exposure and the degree of
leverage employed.
2. Event Driven - A strategy which hopes to benefit from mispricing arising in different
events such as merger arbitrage, restructurings etc. Manager takes a position in an
undervalued security that is anticipated to rise in value because of events such as
mergers, reorganizations, or takeovers. The main risk in such strategies is non-
realization of the event.
3. Equity Hedge - A strategy of investing in equity or equity-like instruments where the
net exposure (gross long minus gross short) is generally low. The manager may invest
globally, or have a more defined geographic, industry or capitalization focus. The risk
primarily pertains to the specific risk of the long and short positions.
4. Restructuring - A strategy of buying and occasionally shorting securities of companies
under Chapter 11 and/or ones which are undergoing some form of reorganization. The
22
securities range from senior secured debt to common stock. The liquidation of
financially distressed company is the main source of risk in these strategies.
5. Event Arbitrage - A strategy of purchasing securities of a company being acquired,
and shorting that of the acquiring company. The risk associated with such strategies is
more of a “deal” risk rather than market risk.
6. Capital Structure Arbitrage - A strategy of buying and selling different securities of the
same issuer (e.g. convertibles/common stock) seeking to obtain low volatility returns
by arbitraging the relative mispricing of these securities.
Directional Strategies: These strategies hope to benefit from broad market movements.
Some popular directional strategies are:
1. Macro - A strategy that seeks to capitalize on country, regional and/or economic
change affecting securities, commodities, interest rates and currency rates. Asset
allocation can be aggressive, and leverage and derivatives may be utilized. The method
and degree of hedging can vary significantly.
2. Long - A strategy which employs a “growth” or “value” approach to investing in
equities with no shorting or hedging to minimize inherent market risk. These funds
mainly invest in the emerging markets where there may be restrictions on short sales.
3. Hedge (Long Bias) - A strategy similar to equity hedge with significant net long
exposure.
4. Short - A strategy that focuses on selling short over-valued securities, with the hope of
repurchasing them in the future at a lower price.
Table I. Sample DescriptionThe table below shows the sample period, total number of funds during the sample period, average number of funds per period and median number of funds perperiod for each of the ten different hedge fund strategies pursued by hedge funds from January 1982 to December 1998.
Quarterly returns Half-Yearly returns Yearly returnsHedge fundstrategy Period Total
fundsDeadfunds
Averagefunds/period
Medianfunds/period
Period Totalfunds
Deadfunds
AverageFunds/perio
d
Medianfunds/period
Period Totalfunds
Deadfunds
Averagefunds/period
Medianfunds/period
Fixed IncomeArbitrage
93Q1-98Q4 25 2 10.3 11.0 93I-98II 22 0 9.8 10.0 93-98 12 0 8.2 9.5
Event Driven 83Q1-98Q4 63 1 21.7 14.0 83I-98II 61 0 21.3 14.0 83-98 56 0 20.4 14.0Equity Hedge 82Q1-98Q4 223 7 57.5 21.0 82I-98II 213 4 56.0 20.5 82-98 174 4 51.5 19.0Restructuring 89Q1-98Q4 34 0 18.0 15.0 89I-98II 34 0 17.7 15.0 89-98 31 0 16.7 14.0Event Arbitrage 84Q1-98Q4 31 2 11.9 9.0 84I-98II 28 0 11.7 9.0 84-98 26 0 11.3 9.0Capital StructureArbitrage
88Q3-98Q4 57 2 21.6 18.0 88II-98II 57 2 25.9 18.0 89-98 37 1 18.9 15.5
Non-Directional 433 14 115.9 53.0 415 6 113.2 51.5 336 5 104.5 47.0Macro 84Q3-98Q4 59 3 19.0 12.5 84II-98II 52 1 18.6 12.0 85-98 38 1 17.6 12.5Long 89Q2-98Q4 41 0 14.0 9.0 89II-98II 40 0 13.8 9.0 90-98 27 0 12.2 9.0Hedge (Long Bias) 82Q2-98Q4 200 10 65.1 45.0 82II-98II 196 8 65.0 45.0 83-98 174 7 63.6 45.5Short 88Q3-98Q4 13 0 6.4 5.0 88II-98II 13 0 6.4 5.0 89-98 11 0 6.3 5.5Directional 313 13 93.7 62.0 301 9 93.3 62.0 250 8 89.8 62.5Overall 746 27 208.2 113.0 716 15 203.7 111.5 586 13 189.0 105.0
Table II. Two-period Performance Persistence of Hedge Fund Strategies on pre-fee and post-fee basis for different return measurement intervals
The table below shows the summary of percentage of cases exhibiting statistically significantpersistence in performance of the ten different hedge fund strategies from January 1982 to December1998. We employ both the parametric (regression-based) and non-parametric (contingency-table-based) methods using Alpha and Appraisal Ratio. The results show the persistence at quarterly, half-yearly and yearly intervals on both pre-fee and post-fee basis . Alpha is defined as the return of thefund manager using a particular strategy minus the average return on all the funds using the samestrategy in that period. Appraisal Ratio is defined as alpha divided by the standard errors of theresiduals from the regression of the fund return on the average return of all the funds following thatstrategy in that period. For contingency table, Winners and Losers are determined by comparing theappraisal ratios of individual fund managers to those of the median manager within each strategy ineach period. WW and LL denote winners and losers in two consecutive periods, LW denotes Losers inthe first period & Winners in the second period & WL denotes the reverse. The Cross-Product Ratio(CPR) and Chi-square statistic computed as per Section III. All figures are in percentage where thefigures in brackets refer to the results on a pre-fee basis.
Panel A: Based on AlphasQuarterly returns Half-Yearly returns Yearly returns
Parametric Non-parametric Parametric Non-parametric Parametric Non-parametricHedge fund strategy
CPR Chi-sq. CPR Chi-sq. CPR Chi-sq.Fixed Income Arb 17 (17) 4 (4) 17 (17) 18 (18) 0 (0) 0 (0) 20 (20) 0 (20) 20 (20)Event Driven 14 (14) 8 (6) 16 (14) 16 (19) 6 (10) 10 (10) 13 (20) 20 (7) 20 (13)Equity Hedge 24 (24) 7 (12) 21 (24) 15 (15) 15 (12) 24 (24) 19 (19) 13 (6) 19 (6)Restructuring 21 (21) 8 (8) 23 (18) 21 (21) 5 (5) 21 (16) 33 (33) 11 (0) 11 (0)Event Arbitrage 5 (7) 0 (0) 22 (25) 7 (7) 0 (0) 17 (14) 0 (0) 0 (0) 0 (0)Capital Structure Arb 27 (29) 7 (5) 12 (10) 40 (40) 20 (25) 25 (35) 22 (22) 22 (22) 33 (33)Non-Directional 31 (33) 21 (24) 31 (36) 15 (15) 15 (18) 21 (24) 19 (19) 25 (19) 31 (31)Macro 11 (11) 4 (4) 18 (18) 14 (14) 0 (0) 11 (11) 8 (8) 15 (0) 23 (8)Long 13 (16) 11 (11) 24 (24) 6 (6) 11 (6) 17 (17) 25 (25) 25 (25) 25 (25)Hedge (Long Bias) 20 (20) 14 (17) 21 (21) 25 (25) 13 (13) 19 (19) 27 (27) 13 (13) 13 (13)Short 7 (7) 0 (0) 22 (24) 10 (10) 0 (5) 20 (20) 0 (0) 0 (0) 33 (33)Directional 27 (27) 18 (21) 26 (27) 38 (38) 13 (16) 16 (22) 27 (33) 20 (20) 20 (21)Overall 34 (34) 24 (27) 34 (36) 30 (33) 27 (30) 36 (42) 25 (25) 25 (19) 31 (25)
Panel B: Based on Appraisal RatiosQuarterly returns Half-Yearly returns Yearly returns
Parametric Non-parametric Parametric Non-parametric Parametric Non-parametricHedge fund strategy
CPR Chi-sq. CPR Chi-sq. CPR Chi-sq.Fixed Income Arb 39 (35) 4 (4) 26 (26) 55 (55) 18 (0) 36 (36) 40 (40) 0 (0) 20 (20)Event Driven 29 (33) 13 (17) 22 (25) 42 (42) 13 (16) 19 (23) 27 (33) 20 (13) 20 (13)Equity Hedge 18 (19) 9 (12) 22 (24) 21 (21) 15 (15) 24 (27) 25 (25) 13 (0) 19 (6)Restructuring 28 (28) 5 (3) 10 (8) 21 (21) 11 (5) 21 (11) 11 (11) 11 (11) 22 (22)Event Arbitrage 8 (9) 2 (2) 25 (25) 3 (10) 0 (0) 17 (17) 0 (0) 0 (0) 0 (0)Capital Structure Arb 32 (34) 10 (10) 17 (12) 35 (35) 20 (35) 35 (45) 33 (22) 22 (22) 44 (44)Non-Directional 55 (57) 25 (24) 36 (33) 39 (39) 15 (24) 27 (33) 25 (25) 19 (18) 44 (38)Macro 28 (28) 9 (7) 23 (23) 32 (32) 0 (7) 18 (18) 38 (46) 8 (8) 15 (15)Long 24 (24) 13 (8) 26 (21) 28 (28) 17 (6) 17 (17) 25 (25) 13 (13) 25 (25)Hedge (Long Bias) 29 (30) 17 (15) 26 (24) 31 (31) 13 (19) 19 (25) 27 (33) 13 (13) 13 (13)Short 12 (15) 2 (2) 34 (34) 10 (10) 0 (10) 25 (35) 22 (22) 0 (0) 22 (33)Directional 41 (45) 29 (27) 41 (38) 44 (44) 13 (16) 19 (19) 33 (40) 20 (20) 20 (21)Overall 51 (52) 33 (33) 42 (42) 45 (45) 27 (27) 36 (42) 38 (38) 19 (19) 25 (25)
Chart I. Theoretical and Observed Frequency Distribution of Series of Wins and Losses of Hedge Fund Strategies using yearly Data
The chart below shows the frequency distribution of a series of wins and losses employing two performance measures, Alphas and Appraisal Ratios, using yearly net-of-feereturns for the ten different strategies pursued by 586 hedge funds from January 1982 to December 1998. Alpha is defined as the return of the fund manager using aparticular strategy minus the average return on all the funds using the same strategy. Appraisal Ratio is defined as alpha divided by the standard errors of the residualsfrom the regression of the fund return on the average return of all the funds following that strategy. Winners and Losers are determined by comparing the alphas andappraisal ratios of individual fund managers to those of the median manager within each strategy in each period.
15
9 N
on-Dir. T
heoretical
Non-D
ir. Winners
Non-D
ir. Losers
Dir. Theoretical
Dir. W
inners
Dir. Losers
Overall T
heoretical
Overall W
inners
Overall Losers
0.00
200.00
400.00
600.00
800.00
1000.00
1200.00
1400.00
1600.00
1800.00
Frequency
No. of Wins /Losses in a row
Distribution Type
Actual and Theoretical Frequency Distributionusing Annual Appraisal Ratios
15
9
Non-D
ir. Theoretical
Non-D
ir. Winners
Non-D
ir. Losers
Dir. Theoretical
Dir. W
inners
Dir. Losers
Overall T
heoretical
Overall W
inners
Overall Losers
0.00
200.00
400.00
600.00
800.00
1000.00
1200.00
1400.00
1600.00
1800.00
Frequency
No. of Wins /Losses in a row
Distribution Type
Actual and Theoretical Frequency Distributionusing Annual Alphas
Table III. Kolmogrov-Smirnov Test for Multi-period Persistence in Performanceof Hedge Fund Strategies
The table below shows the results of the two-sided Kolmogrov-Smirnov test without making any distributionalassumptions about the theoretical distribution of the series of wins and losses for the ten hedge fund strategies. PanelA (B and C) shows the results for the quarterly (half-yearly and yearly) net-of-fee returns of hedge funds from January1982 to December 1998. Bold (Italic) face indicates that the actual distribution of wins/losses is significantly differentfrom the theoretical distribution at 5% (10%) level signifying multi-period persistence in the performance.
Alphas Appraisal RatiosWins Losses Wins LossesHedge Fund Strategy
2-sidedK-S Stat
p-value 2-sidedK-S Stat
p-value 2-sidedK-S Stat
p-value 2-sidedK-S Stat
p-value
Panel A: Based on Quarterly DataFixed Income Arbitrage 0.9806 0.2928 0.6124 0.8475 0.9806 0.29281.2780 0.0763Event Driven 0.5884 0.8793 1.3333 0.0571 1.6973 0.0063 1.6973 0.0063Equity Hedge 0.8575 0.4624 0.8575 0.4624 0.8575 0.4624 0.8575 0.4624Restructuring 0.2357 1.0000 0.4083 0.9963 0.4714 0.9794 0.4083 0.9963Event Arbitrage 0.5884 0.8793 0.5884 0.8793 0.9129 0.3790 0.4083 0.9963Capital Structure Arb 1.0955 0.1815 1.2374 0.0936 1.2057 0.1092 1.2374 0.0936Non-Directional 0.6860 0.7344 1.0000 0.2710 1.2344 0.09501.3887 0.0423Macro 0.2357 1.0000 0.7559 0.6172 1.2005 0.11201.8091 0.0029Long 0.2357 1.0000 0.6396 0.8079 1.0000 0.9674 0.6396 0.8079Hedge (Long Bias) 0.2041 1.0000 0.5477 0.9251 1.0000 0.2710 1.1356 0.1517Short 0.2500 1.0000 0.5000 0.9640 0.2500 1.0000 0.2357 1.0000Directional 0.2041 1.0000 0.7303 0.6604 1.0000 0.27101.5076 0.0212Overall 0.6860 0.7344 0.8333 0.5026 1.2344 0.0950 1.3568 0.0504Panel B: Based on Half-yearly DataFixed Income Arbitrage 0.2673 1.0000 0.5000 0.9640 0.5000 0.9640 0.5345 0.9375Event Driven 0.2357 1.0000 0.4472 0.9883 0.2236 1.0000 0.4714 0.9794Equity Hedge 0.2132 1.0000 0.9129 0.3790 0.2132 1.0000 1.0607 0.2109Restructuring 0.2357 1.0000 0.2500 1.0000 0.2357 1.0000 0.5345 0.9375Event Arbitrage 0.2673 1.0000 1.5435 0.0171 0.6396 0.8079 1.6667 0.0077Capital Structure Arb 0.5000 0.9640 0.6708 0.7591 0.5000 0.9640 0.6708 0.7591Non-Directional 0.2132 1.0000 1.2005 0.1120 0.2132 1.00001.3333 0.0571Macro 0.6124 0.8475 0.6124 0.8475 0.6124 0.8475 0.9806 0.2928Long 0.5774 0.8928 0.6708 0.7591 0.5774 0.8928 0.6708 0.7591Hedge (Long Bias) 0.5000 0.9640 0.4083 0.9963 0.5000 0.9640 0.2236 1.0000Short 0.2673 1.0000 0.6708 0.7591 0.2673 1.0000 0.6708 0.7591Directional 0.2041 1.0000 0.4083 0.9963 0.2041 1.0000 0.5884 0.8793Overall 0.2041 1.0000 1.0290 0.2408 0.2041 1.0000 1.1667 0.1315Panel C: Based on Yearly DataFixed Income Arbitrage 0.3162 1.0000 0.5774 0.8928 0.3162 1.0000 0.5774 0.8928Event Driven 0.4714 0.9794 0.2673 1.0000 0.5774 0.8928 0.2887 1.0000Equity Hedge 0.5345 0.9375 0.8165 0.5320 0.2887 1.0000 0.8165 0.5320Restructuring 0.5000 0.9640 0.3162 1.0000 0.5000 0.9640 0.3162 1.0000Event Arbitrage 0.5774 0.8928 0.4714 0.9794 0.5774 0.8928 0.4714 0.9794Capital Structure Arb 0.3536 1.0000 0.2887 1.0000 0.3536 1.0000 0.3162 1.0000Non-Directional 0.2357 1.0000 0.2041 1.0000 0.7500 0.6272 0.4083 0.9963Macro 0.2500 1.0000 0.2673 1.0000 0.2500 1.0000 0.6708 0.7591Long 0.3162 1.0000 0.2887 1.0000 0.3162 1.0000 0.2887 1.0000Hedge (Long Bias) 0.2887 1.0000 0.4472 0.9883 0.5345 0.9375 0.4472 0.9883Short 0.3162 1.0000 0.8018 0.5587 0.3162 1.0000 0.5345 0.9375Directional 0.7500 0.6272 0.4472 0.9883 0.7500 0.6272 0.2236 1.0000Overall 0.7071 0.6994 0.2041 1.0000 0.7500 0.6272 0.4083 0.9963
Table IV. Kolmogrov-Smirnov Normality Test for Multi-period PersistenceThe table below shows the results of the one-sample Kolmogrov-Smirnov test to compare the observed frequencydistribution of a series of wins & losses for the hedge fund strategies with a normal distribution. Panel A (B &C)shows the results for the quarterly (half-yearly & yearly) net-of-fee returns of hedge funds from January 1982 toDecember 1998. N indicates the total of wins & losses for all the funds following a strategy. Bold (Italic) faceindicates that the observed distribution of wins/losses is significantly different from the normal distribution at 5%(10%) level signifying persistence. Asy. Sig. And MC Sig. stand for asymptotic & Monte-Carlo significancerespectively.
Alphas Appraisal RatiosWins Losses Wins LossesHedge Fund Strategy N
K-SZ-Stat
Asy.Sig.
MCSig.
K-SZ-Stat
Asy.Sig.
MCSig.
K-SZ-Stat
Asy.Sig.
MCSig.
K-SZ-Stat
Asy.Sig.
MCSig.
Panel A: Based on Quarterly DataFixed Income Arb 297 1.02 0.25 0.21 1.04 0.23 0.19 0.95 0.33 0.27 0.97 0.30 0.25Event Driven 1510 1.08 0.19 0.16 1.39 0.04 0.03 1.32 0.06 0.05 1.35 0.05 0.04Equity Hedge 4351 1.48 0.03 0.02 1.42 0.04 0.03 1.45 0.03 0.02 1.40 0.04 0.03Restructuring 837 0.74 0.64 0.56 1.05 0.22 0.18 0.75 0.63 0.55 1.03 0.24 0.20Event Arbitrage 773 1.21 0.11 0.09 1.18 0.13 0.10 1.04 0.23 0.19 0.99 0.29 0.24Capital Structure Arb 1020 1.09 0.19 0.15 1.16 0.14 0.11 1.09 0.19 0.16 1.16 0.14 0.11Non-Directional 8788 1.39 0.04 0.03 1.43 0.03 0.03 1.53 0.02 0.02 1.54 0.02 0.02Macro 1216 0.84 0.48 0.41 1.12 0.16 0.131.32 0.06 0.05 1.56 0.02 0.01Long 627 0.85 0.47 0.39 0.91 0.39 0.32 0.83 0.50 0.43 0.89 0.41 0.35Hedge (Long Bias) 4804 1.02 0.25 0.21 1.26 0.09 0.06 1.35 0.05 0.04 1.44 0.03 0.02Short 296 0.87 0.43 0.35 0.74 0.65 0.56 0.82 0.51 0.43 0.85 0.46 0.39Directional 6943 1.03 0.24 0.20 1.24 0.09 0.07 1.38 0.05 0.04 1.61 0.01 0.01Overall 15731 1.40 0.04 0.03 1.43 0.03 0.03 1.54 0.02 0.02 1.60 0.01 0.01Panel B: Based on Half-yearly DataFixed Income Arb 140 0.67 0.77 0.68 0.63 0.83 0.76 0.70 0.72 0.64 0.55 0.93 0.87Event Driven 755 0.79 0.57 0.50 0.80 0.55 0.48 0.84 0.49 0.42 0.77 0.59 0.52Equity Hedge 2112 1.09 0.18 0.15 1.31 0.06 0.05 1.10 0.18 0.14 1.36 0.05 0.04Restructuring 387 0.89 0.41 0.35 0.78 0.58 0.51 0.84 0.48 0.41 0.68 0.74 0.65Event Arbitrage 379 0.72 0.68 0.59 1.27 0.08 0.06 1.06 0.21 0.17 1.32 0.06 0.05Capital Structure Arb 496 0.71 0.70 0.62 0.80 0.55 0.47 0.70 0.71 0.64 0.76 0.62 0.54Non-Directional 4269 1.04 0.23 0.19 1.36 0.05 0.04 1.02 0.25 0.21 1.42 0.04 0.03Macro 590 1.12 0.17 0.14 0.95 0.32 0.28 1.10 0.18 0.15 0.99 0.28 0.24Long 312 0.68 0.75 0.66 0.88 0.43 0.36 0.67 0.76 0.67 0.85 0.46 0.40Hedge (Long Bias) 2363 0.76 0.60 0.53 1.05 0.22 0.18 0.76 0.61 0.54 0.91 0.38 0.32Short 147 0.64 0.81 0.73 0.88 0.43 0.36 0.62 0.84 0.77 0.81 0.53 0.46Directional 3412 1.12 0.17 0.14 1.03 0.24 0.20 1.11 0.17 0.14 1.11 0.17 0.13Overall 7681 1.11 0.17 0.14 1.38 0.05 0.04 1.10 0.18 0.15 1.44 0.03 0.02Panel C: Based on Yearly DataFixed Income Arb 49 0.56 0.91 0.84 0.56 0.91 0.85 0.56 0.91 0.84 0.56 0.91 0.85Event Driven 326 0.92 0.36 0.30 0.65 0.80 0.72 0.67 0.76 0.67 0.65 0.80 0.72Equity Hedge 879 0.79 0.56 0.47 1.17 0.13 0.10 0.69 0.72 0.63 1.18 0.12 0.10Restructuring 167 0.95 0.33 0.27 0.62 0.84 0.75 0.88 0.43 0.35 0.62 0.84 0.75Event Arbitrage 170 0.70 0.70 0.61 0.81 0.53 0.46 0.73 0.65 0.55 0.85 0.46 0.39Capital Structure Arb 190 0.50 0.97 0.91 0.62 0.84 0.76 0.50 0.97 0.91 0.52 0.95 0.89Non-Directional 1781 0.97 0.31 0.25 1.16 0.14 0.11 0.87 0.44 0.371.20 0.11 0.09Macro 248 0.82 0.52 0.44 0.64 0.81 0.73 0.79 0.57 0.49 0.88 0.42 0.36Long 110 0.55 0.93 0.86 0.59 0.87 0.80 0.51 0.96 0.90 0.54 0.93 0.88Hedge (Long Bias) 1045 0.67 0.77 0.68 1.00 0.28 0.23 0.74 0.64 0.54 1.00 0.27 0.22Short 63 0.49 0.97 0.93 0.61 0.85 0.78 0.54 0.94 0.88 0.70 0.71 0.62Directional 1466 0.81 0.52 0.45 0.98 0.30 0.24 0.79 0.56 0.49 0.98 0.29 0.24Overall 3247 0.95 0.33 0.26 1.15 0.15 0.12 0.83 0.50 0.421.17 0.13 0.10